---
title: "Resilience assessment of NCRMP2024 sites near Guanica"
author: "David Gibbs"
date: "6/11/19"
output:
  word_document: default
  html_document: default
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```



Libraries
```{r necessary libraries}
# library(devtools)
# install_github("vqv/ggbiplot")
library(ggbiplot)
library(xlsx)
library(psych)
library(PerformanceAnalytics)
library(reshape2)
library(stats)
library(psy)
library(BiodiversityR)
library(Metrics)
```



Loads data into R
```{r load data}
#Loads the survey data, calculated as indicators
setwd("C:/Users/dagib/Documents/EPA ORISE EARCG/Coral projects/Adaptation Design Tool/Analyses/Inputs")
all_data <- read.xlsx("NCRMP_2014_sites_SW_PR_20190630.xlsx", sheetIndex=2, header=TRUE)
all_data <- all_data[order(all_data$Station), c(1:9)]

#Converts the values in the indicator spreadheet from factors to numeric
type <- sapply(all_data[,c(2:9)], is.factor)
all_data[type] <- lapply(all_data[type], function(x) as.numeric(as.character(x)))

# Removes rows that have NAs for indicator values
complete_records <- na.omit(all_data)

#Loads the indicator weight spreadsheet
indic_weights <- read.xlsx("NCRMP_2014_sites_SW_PR_20190630.xlsx", sheetIndex = 3, header = TRUE)
indic_weights <- indic_weights[c(1:10),c(3, 5:11)]

#Converts the values in the weighting spreadheet from factors to numeric
weights <- sapply(indic_weights[,-c(1:2)], is.factor)
indic_weights[weights] <- lapply(indic_weights[weights], function(x) as.numeric(as.character(x)))

#Loads the station property data
sta_props <- read.xlsx("NCRMP_PR2014_station_properties.xlsx", sheetIndex = 1, header = TRUE)

#Reorders the station property file by station number
sta_props <- sta_props[order(sta_props$survey_index),]

#Joins the site attributes to the indicators
complete_records_attribs <- merge(x = complete_records, y = sta_props, by.x = "Station", by.y = "survey_index")

#The sites that have coral demographic surveys 
demog_sites_attribs <- complete_records_attribs[which(complete_records_attribs$demo=="1"),]
demog_sites <- demog_sites_attribs[,c(1:9)]

#Just the sites with demographic surveys that actually had full-sized coral colonies.
#This removes the sites that did not have any full-sized coral colonies.
demog_sites_complete_records <- demog_sites_attribs[complete.cases(demog_sites_attribs),]

#Converts the values in the indicator spreadheet from factors to numeric
demog_sites_complete_records[,3] <- as.numeric(as.character(demog_sites_complete_records[,3]))

#Keeps just the demographic survey indicators at the demographic sites
#(coral diversity, coral disease, coral thermal tolerance) 
demog_sites_demog_fields <- demog_sites_complete_records[,c(1:3, 9)]

#All sites without demographic data.
#Removes the columns that are only for sites with demographic data.
all_sites_non_demog_fields <- complete_records_attribs[,c(1, 4:8)]

#Loads the fishing pressure data
# setwd("C:/Users/dagib/Documents/EPA ORISE EARCG/Coral projects/Resilience_assessment/GIS/Stressors/Fishing_pressure")
fishing <- read.csv("NCRMP2014_stations_with_Shivlani_fishing_pressure_sptl_join_20180409.csv", header = TRUE)

#Loads land-based sources of pollution (LBSP) data
# setwd("C:/Users/dagib/Documents/EPA ORISE EARCG/Coral projects/Resilience_assessment/GIS/Stressors/LBSP_coastal_dispersion")
LBSP <- read.xlsx("LBSP_at_NCRMP_sites_90ft_depth_using_7_nearest_OpenNSPECT_pour_points_CCAP2010_20180412.xls", sheetIndex = 10, header = TRUE)

#loads annual bleaching onset year (scenario rcp8.5) estimates for each site
# setwd("C:/Users/dagib/Documents/EPA ORISE EARCG/Coral projects/Resilience_assessment/GIS/UNEP van Hooindonk 2016 bleaching projections")
annual_bleaching_rcp85 <- read.csv("annual_bleaching_onset_year_rcp85_20180424.csv", header = TRUE)

```



Performs principal components analysis as shown at
https://www.r-bloggers.com/computing-and-visualizing-pca-in-r/
using the prcomp method
```{r performs PCA}

#PCA for sites with demographic data using demog and non-demog (i.e. all) indicators
demog_sites_pca <- prcomp(demog_sites_complete_records[,c(2:9)], center=TRUE, scale.=TRUE)
print(demog_sites_pca)
summary(demog_sites_pca)
plot(demog_sites_pca, type = "lines")
#Matrix of each site with its PCA scores
demog_sites_pca_scores <- demog_sites_pca$x
demog_sites_pca_scores <- cbind(demog_sites_complete_records[,1], demog_sites_pca_scores)

# #PCA for all sites using non-demographic indicators only
# all_sites_non_demog_fields_pca <- prcomp(all_sites_non_demog_fields[,c(2:5)], center=TRUE, scale.=TRUE)
# print(all_sites_non_demog_fields_pca)
# summary(all_sites_non_demog_fields_pca)
# plot(all_sites_non_demog_fields_pca, type = "lines")
# #Matrix of each site with its PCA scores
# all_sites_non_demog_fields_pca_scores <- all_sites_non_demog_fields_pca$x
# all_sites_non_demog_fields_pca_scores <- cbind(all_sites_non_demog_fields[,1], all_sites_non_demog_fields_pca_scores)
# 
# #PCA for demog sites using demographic indicators only at demographic sites
# demog_sites_demog_fields_pca <- prcomp(demog_sites_demog_fields[,c(2:4)], center=TRUE, scale.=TRUE)
# print(demog_sites_demog_fields_pca)
# summary(demog_sites_demog_fields_pca)
# plot(demog_sites_demog_fields_pca, type = "lines")
# #Matrix of each site with its PCA scores
# demog_sites_demog_fields_pca_scores <- demog_sites_demog_fields_pca$x
# demog_sites_demog_fields_pca_scores <- cbind(demog_sites_demog_fields[,1], demog_sites_demog_fields_pca_scores)

```



Creates PCA ordination plots
```{r PCA ordination plots}

#PCA ordination plots for sites with demographic data using demog and non-demog (i.e. all) indicators
demog_sites_pca_plot <- ggbiplot(demog_sites_pca, obs.scale = 1, var.scale = 1)
demog_sites_pca_plot <- demog_sites_pca_plot + scale_color_discrete(name = '')
demog_sites_pca_plot <- demog_sites_pca_plot + theme(legend.direction = 'horizontal', 
               legend.position = 'top')
demog_sites_pca_plot <- demog_sites_pca_plot + ggtitle("PCA for sites with demographic data")
print(demog_sites_pca_plot)

# #PCA ordination plots for demographic indicators at demographic sites
# demog_pca_plot <- ggbiplot(demog_sites_demog_fields_pca, obs.scale = 1, var.scale = 1)
# demog_pca_plot <- demog_pca_plot + scale_color_discrete(name = '')
# demog_pca_plot <- demog_pca_plot + theme(legend.direction = 'horizontal', 
#                legend.position = 'top')
# demog_pca_plot <- demog_pca_plot + ggtitle("PCA for demographic indicators at sites with demographic data")
# print(demog_pca_plot)
# 
# #PCA ordination plots for all sites using non-demographic indicators only
# non_demog_pca_plot <- ggbiplot(all_sites_non_demog_fields_pca, obs.scale = 1, var.scale = 1)
# non_demog_pca_plot <- non_demog_pca_plot + scale_color_discrete(name = '')
# non_demog_pca_plot <- non_demog_pca_plot + theme(legend.direction = 'horizontal', 
#                legend.position = 'top')
# non_demog_pca_plot <- non_demog_pca_plot + ggtitle("PCA for non-demographic indicators at all sites")
# print(non_demog_pca_plot)

```



Correlations between all pairs of indicators
```{r Correlations between indicators}
#Sites with coral demographic data that have all indicator values (i.e. excludes 8 sites that did not have any demogrphic data)
chart.Correlation(demog_sites_complete_records[,c(2:9)], histogram=TRUE, method=c("spearman"), pch=".")
demog_indic_cor <- cor(demog_sites_complete_records[,c(2:9)], use="pairwise.complete.obs", method=c("spearman"))
# pairs.panels(demog_sites_complete_records[,c(2:9)], method="spearman", lm=TRUE, ellipses=FALSE)

# #Sites with demographic data but just using the indicators that don't come from the demographic surveys
# chart.Correlation(all_sites_non_demog_fields[,c(2:6)], histogram=TRUE, method=c("spearman"), pch=".")
# # pairs.panels(all_sites_non_demog_fields[,c(2:6)], method="spearman", lm=TRUE, ellipses=FALSE)

write.table(demog_indic_cor, file = paste("demog_indic_Pearson_", Sys.Date(), ".csv", sep=""), row.names = FALSE)
```


Factor analysis between all indicators
Sources include: 
http://www.statpower.net/Content/312/R%20Stuff/Exploratory%20Factor%20Analysis%20with%20R.pdf
https://www.promptcloud.com/blog/exploratory-factor-analysis-in-r/
http://www.di.fc.ul.pt/~jpn/r/factoranalysis/factoranalysis.html#introduction
```{r Factor analysis between all indicators}

#Conducts the factor analysis using varimax rotation
demog_fa <- factanal(demog_sites_complete_records[,c(2:9)], factors=4, rotation="varimax")
demog_fa

#Scree plot of factors' importance to show how many factors to use
scree.plot(demog_fa$correlation)

#Plots the loadings of the indicators on two of the factors
demog_fa_load <- demog_fa$loadings[,c(1:2)]
plot(demog_fa_load, type="n")
text(demog_fa_load, labels=names(demog_sites_complete_records[,c(2:9)]), cex=0.7)

```



Rescales indicators
```{r Rescale indicators}

#Divides each indicator by the largest value for that indicator across all sites with demographic data
demog_sites_rescaled <- t(t(demog_sites_complete_records[,c(2:9)])/apply(demog_sites_complete_records[,c(2:9)], 2, function(x) max(na.omit(x))))

#Binds the rescaled indicators to the raw indicators
demog_sites_rescaled <- cbind(demog_sites_complete_records[,c(1:9)], demog_sites_rescaled)

#Renames the columns to distinguish the raw indicators from the rescaled indicators
colnames(demog_sites_rescaled)[1] <- colnames(demog_sites_attribs)[1]
colnames(demog_sites_rescaled)[c(2:9)] <- paste(colnames(demog_sites_rescaled)[c(2:9)], "_raw", sep="")
colnames(demog_sites_rescaled)[c(10:17)] <- paste(colnames(demog_sites_rescaled)[c(10:17)], "_rescaled", sep="")

# #Turns unweighted indicators into long form for box plot. Also reorders the variables for box plot.
# demog_sites_rescaled_long <- melt(demog_sites_rescaled[,c(10:17)])
# demog_sites_rescaled_long$variable <- factor(demog_sites_rescaled_long$variable, levels=c("SimpsonDiv_rescaled", "FractNotDiseased_rescaled","ThermalTol_rescaled", "HardCoralCover_rescaled", "AlgaeCover_rescaled", "Rugosity_rescaled", "HerbivoreBiomass_rescaled", "TempVar_rescaled"))
# 
# #creates boxplot of each rescaled indicator (1 is maximum value)
# p <- ggplot(data=demog_sites_rescaled_long, aes(x=variable, y=value)) + geom_boxplot(aes(fill=variable))
# p <- p + theme(legend.position="none")
# p <- p + theme(axis.text.x = element_text(angle = 65, vjust = 0.6))
# p <- p + labs(x = "Indicator", y = "Rescaled value")
# p <- p + scale_fill_manual(values=c("#999555", "#999555", "#999555", "#999555", "#999555", "#999555", "#999555", "#999555"))
# p

```



Functions to calculate composite resilience scores, ranks, and rank quartiles at sites with demographic data: 1) using all indicators, and 2) without the temperature variation indicator
```{r Functions to calculate unweighted resilience scores}

#Calculates the composite resilience score by averaging the rescaled indicators
resilScore <- function(rescaled_indics) {
  
  #Creates empty fields for the resilience scores
  rescaled_indics$resil_all_indic <- NA
  rescaled_indics$resil_no_temp_var <- NA
  
  #Calculates resilience scores with all indicators and without the temperature variation indicator.
  resil_all_indic_demog <- rowMeans(rescaled_indics[which(rescaled_indics$demo == 1), c(10:17)], na.rm = TRUE)
  resil_no_temp_var_demog <- rowMeans(rescaled_indics[which(rescaled_indics$demo == 1), c(10:15,17)], na.rm = TRUE)
  
  #Rescales the resilience scores to the highest resilience score
  resil_all_indic_demog <- resil_all_indic_demog/max(resil_all_indic_demog)
  resil_no_temp_var_demog <- resil_no_temp_var_demog/max(resil_no_temp_var_demog)
  
  #Attaches the resilience scores (all indicators and without temperature variation) to the rescaled indicator table
  rescaled_indics[which(rescaled_indics$demo == 1), ]$resil_all_indic <- resil_all_indic_demog
  rescaled_indics[which(rescaled_indics$demo == 1), ]$resil_no_temp_var <- resil_no_temp_var_demog
  
  return(rescaled_indics)
}

#Calculates the resilience score rank (1 is highest resilience) for each site
resilRank <- function(rescaled_indics) {
  
  rescaled_indics <- as.data.frame(rescaled_indics)
  
  #Creates empty columns for the resilience ranks
  rescaled_indics$resil_rank_all_indic <- NA
  rescaled_indics$resil_rank_no_temp_var <- NA
  
  #Resilience ranks for demographic sites using all indicators
  rescaled_indics[which(rescaled_indics$demo == 1),]$resil_rank_all_indic <- rank(-rescaled_indics[which(rescaled_indics$demo == 1),]$resil_all_indic, na.last = "keep", ties.method = "first")

  #Resilience ranks for demographic sites using all indicators except temperature variation
  rescaled_indics[which(rescaled_indics$demo == 1),]$resil_rank_no_temp_var <- rank(-rescaled_indics[which(rescaled_indics$demo == 1),]$resil_no_temp_var, na.last = "keep", ties.method = "first")
  
  return(rescaled_indics)
}

#Calculates the quartile of resilience that each site is in (1 is most resilient quartile, 4 is least resilient quartile
resilQuartile <- function(rescaled_indics) {
  
  #Creates empty columns for the resilience rank quartiles
  rescaled_indics$resil_quart_all_indic <- NA
  rescaled_indics$resil_quart_no_temp_var <- NA
  
  #Resilience rank quartiles for demographic sites using all indicators
  rescaled_indics[which(rescaled_indics$demo == 1),]$resil_quart_all_indic <-
    cut(rescaled_indics[which(rescaled_indics$demo == 1),]$resil_rank_all_indic,        breaks=quantile(rescaled_indics[which(rescaled_indics$demo == 1),]$resil_rank_all_indic, probs=seq(0, 1, 0.25)), include.lowest=TRUE, labels = 1:4)
  
  # #Resilience rank quartiless for demographic sites using all indicators except temperature variation
  rescaled_indics[which(rescaled_indics$demo == 1),]$resil_quart_no_temp_var <-
    cut(rescaled_indics[which(rescaled_indics$demo == 1),]$resil_rank_no_temp_var,        breaks=quantile(rescaled_indics[which(rescaled_indics$demo == 1),]$resil_rank_no_temp_var, probs=seq(0, 1, 0.25)), include.lowest=TRUE, labels = 1:4)

  return(rescaled_indics)
}

```



Calculates resilience scores, ranks, and quartiles for unweighted indicators
```{r Calculates resilience scores, ranks, and quartiles for unweighted indicators}

demog_sites_rescaled <- as.data.frame(demog_sites_rescaled)

#Attaches the site attributes to the indicators and resilience scores
demog_sites_rescaled_attribs <- merge(x = demog_sites_rescaled, y = sta_props, by.x = "Station", by.y = "survey_index")

#Resilience scores based on rescaled indicators
demog_sites_rescaled_scores <- resilScore(demog_sites_rescaled_attribs)

#Resilience ranks based on the composite resilience scores
demog_sites_rescaled_ranks <- resilRank(demog_sites_rescaled_scores)

#Resilience quartiles based on the resilience ranks (1 is most resilient quartile, 4 is least resilient quartile)
demog_sites_rescaled_quartiles <- resilQuartile(demog_sites_rescaled_ranks)

#Adds a field with the weighting system for the indicators. This one is unweighted.
demog_sites_rescaled_quartiles["weighting_sys"] <- 0

#Deletes a redundant field
demog_sites_rescaled_quartiles <- demog_sites_rescaled_quartiles[,-c(21)]

#Adds four fields for comparing the unweighted and weighted ranks and quartiles (two for ranks/quartiles with all indicators and two for ranks/quartiles without temperature variation). Does not apply for the unweighted data, of course.
demog_sites_rescaled_quartiles["quartile_change_all_indic"] <- "N/A"
demog_sites_rescaled_quartiles["diff_unweighted_weighted_all_indic"] <- "N/A"
demog_sites_rescaled_quartiles["quartile_change_no_temp_var"] <- "N/A"
demog_sites_rescaled_quartiles["diff_unweighted_weighted_no_temp_var"] <- "N/A"

#Saves the unweighted data in its own dataframe
data_demog_sites_unweighted <- demog_sites_rescaled_quartiles

write.table(data_demog_sites_unweighted, file = paste("demog_sites_unweighted_", Sys.Date(), ".csv", sep=""), row.names = FALSE)

```



Calculates resilience scores for each site using different weighting systems
```{r Calculates resilience scores using different weighting systems}

#Creates a new dataframe for weighted data
data_demog_sites_weighted <- demog_sites_rescaled_quartiles

for(i in 2:nrow(indic_weights)) {

  #Creates an empty weighting table
  weighted <- NA

  #Iterates through every station
  for(j in 1:nrow(data_demog_sites_unweighted)) {
  
    #Multiplies the rescaled values by the indicator weights
    weighted_new <- data_demog_sites_unweighted[j,c(10:17)]*indic_weights[i,c(1:8)]
    weighted <- rbind(weighted, weighted_new)
   
  }

  #Removes the empty first row of the weighted table
  weighted <- weighted[-c(1),]

  #Adds the station property columns to the weighted values and rearranges columns so they match the unweighted table
  weighted <- cbind(data_demog_sites_weighted[c(1:nrow(data_demog_sites_unweighted)),c(1:9)], weighted)
  weighted <- cbind(weighted, data_demog_sites_weighted[c(1:nrow(data_demog_sites_unweighted)),c(18:ncol(data_demog_sites_weighted))])

  #Resilience scores based on all rescaled indicators available at a site
  weighted <- resilScore(weighted)

  #Resilience ranks (separate for sites with and without demographic indicators (111 and 119, respectively))
  weighted <- resilRank(weighted)

  #Resilience quartiles (separate for sites with and without demographic indicators (111 and 119, respectively))
  weighted <- resilQuartile(weighted)

  #Adds a field with the weighting system for the indicators
  weighted["weighting_sys"] <- i-1

  #Adds four fields for comparing the unweighted and weighted ranks and quartiles (two for ranks/quartiles with all indicators and two for ranks/quartiles without temperature variation). Does not apply for the unweighted data, of course.
  weighted["quartile_change_all_indic"] <- NA
  weighted["diff_unweighted_weighted_all_indic"] <- NA
  weighted["quartile_change_no_temp_var"] <- NA
  weighted["diff_unweighted_weighted_no_temp_var"] <- NA

  #Adds and populates columns for storing changes in resilience rank and quartile between the unweighted and weighted systems
  weighted$quartile_change_all_indic <- paste(data_demog_sites_unweighted$resil_quart_all_indic, " to ", weighted$resil_quart_all_indic)
  weighted$diff_unweighted_weighted_all_indic <- data_demog_sites_unweighted$resil_rank_all_indic - weighted$resil_rank_all_indic
  weighted$quartile_change_no_temp_var <- paste(data_demog_sites_unweighted$resil_quart_no_temp_var, " to ", weighted$resil_quart_no_temp_var)
  weighted$diff_unweighted_weighted_no_temp_var <- data_demog_sites_unweighted$resil_rank_no_temp_var - weighted$resil_rank_no_temp_var

  #Gives the weighted table the same column names as the unweighted table
  colnames(weighted) <- colnames(data_demog_sites_weighted)

  #Adds the weighted table to the end of the previous table
  data_demog_sites_weighted <- rbind(data_demog_sites_weighted, weighted)

}

write.table(data_demog_sites_weighted, file = paste("resil_with_weighting_", Sys.Date(), ".csv", sep=""), row.names = FALSE)

```



Plots comparing ranks of unweighted resilience ranks and some of the weighted resilience ranks against each other
Also calculates the root mean square error of ranks under the weighting systems against the ranks under the unweighted system
```{r Plots comparing ranks of unweighted resilience ranks and some of the weighted resilience ranks against each other. Also, calculated RMSE of weighted vs. unweighted.}

#Function to plot the resilience score ranks. Run separately on sites with and without demographic indicators. This is just for resilience ranks calculated without the temperature variation indicator.
resilRankCompare <- function(sensit_table, demog) {
  
  #Table of unweighted ranks
  unweighted_table <- sensit_table[which(sensit_table$demo==demog & sensit_table$weighting_sys == 0),]$resil_rank_no_temp_var

  rank_table <- data.frame(unweighted_table)
  
  colnames(rank_table) <- "Unweighted"

  #Iterates through the chosen weighting systems
  for (i in c(3,7,9)){

    #For each weighting system, a dataframe with the ranks
    weighted_table <- sensit_table[which(sensit_table$demo==demog & sensit_table$weighting_sys == i),]$resil_rank_no_temp_var
  
    weighted_table <- data.frame(weighted_table)
  
    colnames(weighted_table) <- paste("Weighting_", i, sep="")

    #Attaches the table with weighted ranks
    rank_table <- cbind(rank_table, weighted_table)

  }

  #Creates the plot
  g_total<- ggplot(rank_table, aes(x = Unweighted, y = Weighted_7, color = Weighting)) + 
    geom_point(aes(y = rank_table[,2], col = gsub("_", " ", colnames(rank_table)[2])), shape=19, size=3) + 
    geom_point(aes(y = rank_table[,3], col = gsub("_", " ", colnames(rank_table)[3])), shape=15, size=3) + 
    geom_point(aes(y = rank_table[,4], col = gsub("_", " ", colnames(rank_table)[4])), shape=18, size=3) +
    scale_color_grey(start=0.65, end=0.05)
  
  g_total <- g_total + labs(x = "Unweighted rank", y = "Weighted rank", color = "Weighting system")
  g_total <- g_total + xlim(0, nrow(rank_table)+10)
  g_total <- g_total + ylim(0, nrow(rank_table)+10)
  # g_total <- g_total + ggtitle("Weighted vs. unweighted resilience score ranks")
  # g_total <- g_total + labs(subtitle = "For select weighting systems")
  g_total <- g_total + theme(plot.title = element_text(hjust = 0.5))
  g_total <- g_total + theme(plot.subtitle=element_text(hjust = 0.5))
  g_total <- g_total + scale_x_continuous(breaks=c(seq(0,nrow(rank_table)+10,25)))
  g_total <- g_total + scale_y_continuous(breaks=c(seq(0,nrow(rank_table)+10,25)))
  g_total <- g_total + theme_bw()
  g_total <- g_total + guides(color = guide_legend(override.aes = list(shape = c(19,15,18))))
  g_total <- g_total + theme(legend.position = c(0.9,0.2))

  # #Different subtitles depending on whether demographic sites or non-demographic sites are being plotted
  # if (demog==0) {
  #   g_total<- g_total+ labs(subtitle = "Only sites without demographic indicators")
  # }
  # if (demog==1) {
  #   g_total<- g_total+ labs(subtitle = "Only sites with demographic data (no temperature variation indicator)")
  # }
  
  return(g_total)
  
}

# p1 <- resilRankCompare(all_data_rescaled_attribs, 0)
# p1

p2 <- resilRankCompare(data_demog_sites_weighted, 1)
p2

ggsave("indic_sensit_analysis.jpg", plot = p2, device = "jpeg", width=10, height=8, dpi=300)

####Calculates root mean square error (RMSE)for each comparison of weighted and unweighted indicator rankings

#Creates an empty dataframe
rmse_weights <- data.frame(matrix(NA, nrow=9, ncol=1))
colnames(rmse_weights) <- "RMSE"

#Unweighted ranks
unweighted_ranks <- data_demog_sites_weighted[which(data_demog_sites_weighted$weighting_sys == 0),]$resil_rank_no_temp_var

#Calculates RMSE for each weighted ranking system
for (i in 1:max(data_demog_sites_weighted$weighting_sys)){
  
  weighted_ranks <- data_demog_sites_weighted[which(data_demog_sites_weighted$weighting_sys == i),]$resil_rank_no_temp_var
  
  rmse_weights[i,1] <- rmse(unweighted_ranks, weighted_ranks)
  
}

print(rmse_weights)

```



Plots comparing the difference between weighted resilience ranks against unweighted ranks
```{r Plots comparing the difference between weighted resilience ranks against unweighted ranks}

#Function to plot the resilience score ranks. Run separately on sites with and without demographic indicators. This is just for resilience ranks calculated without the temperature variation indicator.
resilRankDiff <- function(sensit_table, demog) {
  
  #Table of unweighted ranks
  unweighted_table <- sensit_table[which(sensit_table$demo==demog & sensit_table$weighting_sys == 0),]$resil_rank_no_temp_var

  rank_table <- data.frame(unweighted_table)
  
  colnames(rank_table) <- "Unweighted"

  #Iterates through the chosen weighting systems
  for (i in c(3,7,9)){

    #For each weighting system, a dataframe with the rank differences
    weighted_table <- sensit_table[which(sensit_table$demo==demog & sensit_table$weighting_sys == i),]$diff_unweighted_weighted_no_temp_var
  
    weighted_table <- data.frame(weighted_table)
  
    colnames(weighted_table) <- paste("Weighting_", i, sep="")

    #Attaches the table with weighted ranks
    rank_table <- cbind(rank_table, weighted_table)

  }
  
  #Converts the rank differences into numbers
  rank_table[,c(2:4)] <- lapply(rank_table[,c(2:4)], function(x) as.numeric(as.character(x)))

  #Creates the plot
  g_diff<- ggplot(rank_table, aes(x = Unweighted, y = Weighted_7, color = Weighting)) + 
    geom_point(aes(y = rank_table[,2], col = colnames(rank_table)[2])) + 
    geom_point(aes(y = rank_table[,3], col = colnames(rank_table)[3])) + 
    geom_point(aes(y = rank_table[,4], col = colnames(rank_table)[4]))
  
  g_diff<- g_diff+ labs(x = "Unweighted rank", y = "Difference from unweighted rank")
  g_diff<- g_diff+ ggtitle("Difference between weighted and unweighted ranks")
  g_diff<- g_diff+ theme(plot.title = element_text(hjust = 0.5))
  g_diff<- g_diff+ theme(plot.subtitle = element_text(hjust = 0.5))
  g_diff <- g_diff + theme_bw()
  
  #Different subtitles depending on whether demographic sites or non-demographic sites are being plotted
  if (demog==0) {
    g_diff<- g_diff+ labs(subtitle = "Only sites without demographic indicators")
  }
  if (demog==1) {
    g_diff<- g_diff+ labs(subtitle = "Only sites with demographic data (no temperature variation indicator)")
  }
  
  return(g_diff)
  
}

# p3 <- resilRankCompare(all_data_rescaled_attribs, 0)
# p3

p4 <- resilRankDiff(data_demog_sites_weighted, 1)
p4

```



Matrix of how many sites change quartiles using different weighting systems. Only for quartiles assigned without the temperature variation indicator.
```{r Table of how many sites change quartiles using different weighting systems}

#Makes the quartile change field a factor
data_demog_sites_weighted$quartile_change_no_temp_var <- as.factor(data_demog_sites_weighted$quartile_change_no_temp_var)

#dataframe for quartile changes that is one column for each weighting scheme and one row for each quartile change found in the data
quartile_change <- data.frame(matrix(0, nrow =  nlevels(data_demog_sites_weighted$quartile_change_no_temp_var), ncol = max(data_demog_sites_weighted$weighting_sys)))

#Assigns row and column names
row.names(quartile_change) <- c(levels(data_demog_sites_weighted$quartile_change_no_temp_var))
colnames(quartile_change) <- paste("Weighting_", c(1:max(data_demog_sites_weighted$weighting_sys)), sep="")

#Deletes the empty first row
quartile_change <- quartile_change[-c(nrow(quartile_change)),]

#Iteratres through all the weighting schemes (i.e. columns of the quartile_change dataframe)
for (i in 1:max(data_demog_sites_weighted$weighting_sys)) {
  
  #Pulls out just the appropriate weighting
  weighted <- data_demog_sites_weighted[which(data_demog_sites_weighted$weighting_sys == i),]

  #Iterates through every row (i.e. station) of the weighting to pull out its quartile transition
  for (j in 1:nrow(weighted)) {

    #The actual transition for the row (station) of the weighting system that is being added to the transition matrix
    change <- weighted$quartile_change_no_temp_var[j]

    #Increments the value in the transition matrix by 1 for each transition found
    value <- as.integer(subset(quartile_change, rownames(quartile_change) %in% change)[i]) + 1

    quartile_change[which(rownames(quartile_change) %in% change),i] <- value

  }
}

#Sums the number of transitions for each weighting system. It should be 230.
quartile_change["Sum",] <- colSums(quartile_change)

quartile_change <- quartile_change/nrow(data_demog_sites_unweighted)

print(quartile_change)

write.table(quartile_change, file = paste("sensit_analysis_quartile_change_", Sys.Date(), ".csv", sep=""), row.names = TRUE)

```



Merges LBSP and fishing stressor data with indicator data and makes boxplots of rescaled stressors and indicators
```{r Merges LBSP and fishing stressor data with indicator data and makes boxplots}

#Calculates the average of the rescaled nitrogen and sediment values for each site
LBSP$StrdLBSP <- rowMeans(LBSP[,c("StrdNtrgen", "StrdSedmnt")])

#Replaces the non-existent LBSP values with NAs
LBSP[LBSP$StrdNtrgen == -99999, ]$StrdNtrgen <- NA
LBSP[LBSP$StrdSedmnt == -99999, ]$StrdSedmnt <- NA
LBSP[LBSP$StrdLBSP == -99999, ]$StrdLBSP <- NA

#Only the necessary columns from the fishing data, and changes field names
fishing_no_attribs <- fishing[,c(4, 32:36)]
colnames(fishing_no_attribs)[c(2:6)] <- paste(colnames(fishing_no_attribs)[c(2:6)], "_fishing", sep="")

#New dataset that will have indicators and stressors
demog_sites_with_stressors <- data_demog_sites_weighted

#Merges the fishing data to the indicator data
demog_sites_with_stressors <- merge(x = demog_sites_with_stressors, y = fishing_no_attribs, by.x = "Station", by.y = "SurveyIndx")

#Merges the LBSP data to the indicator data
demog_sites_with_stressors <- merge(x = demog_sites_with_stressors, y = LBSP[,-c(1)], by.x = "Station", by.y = "StationID")

#Rescales fishing values to a highest value of 1 for the sites in the assessment
demog_sites_with_stressors$all_total_fishing_rescaled  <- demog_sites_with_stressors$all_total_fishing/max(demog_sites_with_stressors$all_total_fishing, na.rm=TRUE)

#Rescales the LBSP values to a highest value of 1 for the sites in the assessment
demog_sites_with_stressors$StrdNtrgen <- demog_sites_with_stressors$StrdNtrgen/max(demog_sites_with_stressors$StrdNtrgen, na.rm=TRUE)
demog_sites_with_stressors$StrdSedmnt <- demog_sites_with_stressors$StrdSedmnt/max(demog_sites_with_stressors$StrdSedmnt, na.rm=TRUE)
demog_sites_with_stressors$StrdLBSP <- demog_sites_with_stressors$StrdLBSP/max(demog_sites_with_stressors$StrdLBSP, na.rm=TRUE)

#Since the above merges changed the order of the rows, reorders the indicator/stressor data by the weighting system, then by the station ID
demog_sites_with_stressors <- demog_sites_with_stressors[order(demog_sites_with_stressors$weighting_sys, demog_sites_with_stressors$Station),]

#Turns unweighted indicators and stressors into long form for box plot. Also reorders the variables for box plot.
demog_sites_rescaled_long <- melt(demog_sites_with_stressors[which(demog_sites_with_stressors$weighting_sys == 0),c(10:17, 41,42, 64:67)])
demog_sites_rescaled_long$variable <- factor(demog_sites_rescaled_long$variable, levels=c("SimpsonDiv_rescaled", "FractNotDiseased_rescaled","ThermalTol_rescaled", "HardCoralCover_rescaled", "AlgaeCover_rescaled", "Rugosity_rescaled", "HerbivoreBiomass_rescaled", "TempVar_rescaled", "resil_all_indic", "resil_no_temp_var", "StrdNtrgen", "StrdSedmnt", "StrdLBSP", "all_total_fishing_rescaled"))

#creates boxplot of each rescaled indicator (1 is maximum value)
p <- ggplot(data=demog_sites_rescaled_long, aes(x=variable, y=value)) + geom_boxplot(aes(fill=variable))
p <- p + theme_bw()
p <- p + theme(text = element_text(size=14), 
               axis.text.x = element_text(angle = 0, vjust = 0.6))
p <- p + labs(x = "Metric \n (indicator, stressor, or resilience score)", y = "Rescaled value")
p <- p + scale_fill_manual(values=c("#999555", "#999555", "#999555", "#999555", "#999555", "#999555", "#999555", "#999555", "blue", "blue", "#999999", "#999999", "#999999", "#999999"))
p <- p + scale_x_discrete(labels=c("Simpson \n diversity ", "Colonies \n not \n diseased","Coral \n thermal \n tolerance", "Hard \n coral \n cover", "Macroalgae \n cover", "Rugosity", "Herbivore \n biomass", "Temperature \n variation", "Resilience- \n all indicators", "Resilience- \n without \n temperature \n variation", "Nitrogen \n level", "Sediment \n level", "N + sed \n average", "Fishing \n pressure"))
p <- p + scale_y_continuous(breaks=c(seq(0,1,0.2)))
p <- p + theme(legend.position="none")

#Minimum, mean, and maximum values of raw indicators, where appropriate
simpson_min <- min(data_demog_sites_unweighted$SimpsonDiv_raw)
simpson_mean <- mean(data_demog_sites_unweighted$SimpsonDiv_raw)
simpson_max <- max(data_demog_sites_unweighted$SimpsonDiv_raw)
disease_min <- min(data_demog_sites_unweighted$FractNotDiseased_raw)
disease_mean <- mean(data_demog_sites_unweighted$FractNotDiseased_raw)
disease_max <- max(data_demog_sites_unweighted$FractNotDiseased_raw)
thermalTol_min <- min(data_demog_sites_unweighted$ThermalTol_raw)
thermalTol_mean <- mean(data_demog_sites_unweighted$ThermalTol_raw)
thermalTol_max <- max(data_demog_sites_unweighted$ThermalTol_raw)
coralCover_min <- min(data_demog_sites_unweighted$HardCoralCover_raw)
coralCover_mean <- mean(data_demog_sites_unweighted$HardCoralCover_raw)
coralCover_max <- max(data_demog_sites_unweighted$HardCoralCover_raw)
algaeCover_min <- min(data_demog_sites_unweighted$AlgaeCover_raw)
algaeCover_mean <- mean(data_demog_sites_unweighted$AlgaeCover_raw)
algaeCover_max <- max(data_demog_sites_unweighted$AlgaeCover_raw)
rugosity_min <- min(data_demog_sites_unweighted$Rugosity_raw)
rugosity_mean <- mean(data_demog_sites_unweighted$Rugosity_raw)
rugosity_max <- max(data_demog_sites_unweighted$Rugosity_raw)
herbiv_min <- min(data_demog_sites_unweighted$HerbivoreBiomass_raw)
herbiv_mean <- mean(data_demog_sites_unweighted$HerbivoreBiomass_raw)
herbiv_max <- max(data_demog_sites_unweighted$HerbivoreBiomass_raw)
tempVar_min <- min(data_demog_sites_unweighted$TempVar_raw)
tempVar_mean <- mean(data_demog_sites_unweighted$TempVar_raw)
tempVar_max <- max(data_demog_sites_unweighted$TempVar_raw)

#Adds minimum, mean and maximum values of raw indicators, where applicable
p <- p + annotate("text", x=0.6, y=1.1, label="Min", size = 4)
p <- p + annotate("text", x=0.6, y=1.15, label="Mean", size = 4)
p <- p + annotate("text", x=0.6, y=1.2, label="Max", size = 4)
p <- p + annotate("text", x=1, y=1.1, label=round(simpson_min, 2), size = 4)
p <- p + annotate("text", x=1, y=1.15, label=round(simpson_mean, 2), size = 4)
p <- p + annotate("text", x=1, y=1.2, label=round(simpson_max, 2), size = 4)
p <- p + annotate("text", x=2, y=1.1, label=paste(round(disease_min*100, 0), "%", sep=""), size = 4)
p <- p + annotate("text", x=2, y=1.15, label=paste(round(disease_mean*100, 0), "%", sep=""), size = 4)
p <- p + annotate("text", x=2, y=1.2, label=paste(round(disease_max*100, 0), "%", sep=""), size = 4)
p <- p + annotate("text", x=3, y=1.1, label=round(thermalTol_min, 2), size = 4)
p <- p + annotate("text", x=3, y=1.15, label=round(thermalTol_mean, 2), size = 4)
p <- p + annotate("text", x=3, y=1.2, label=round(thermalTol_max, 2), size = 4)
p <- p + annotate("text", x=4, y=1.1, label=paste(round(coralCover_min, 0), "%", sep=""), size = 4)
p <- p + annotate("text", x=4, y=1.15, label=paste(round(coralCover_mean, 0), "%", sep=""), size = 4)
p <- p + annotate("text", x=4, y=1.2, label=paste(round(coralCover_max, 0), "%", sep=""), size = 4)
p <- p + annotate("text", x=5, y=1.1, label=paste(round(algaeCover_min, 0), "%", sep=""), size = 4)
p <- p + annotate("text", x=5, y=1.15, label=paste(round(algaeCover_mean, 0), "%", sep=""), size = 4)
p <- p + annotate("text", x=5, y=1.2, label=paste(round(algaeCover_max, 0), "%", sep=""), size = 4)
p <- p + annotate("text", x=6, y=1.1, label=round(rugosity_min, 2), size = 4)
p <- p + annotate("text", x=6, y=1.15, label=round(rugosity_mean, 2), size = 4)
p <- p + annotate("text", x=6, y=1.2, label=round(rugosity_max, 2), size = 4)
p <- p + annotate("text", x=7, y=1.08, label=paste(round(herbiv_min, 2), "\n g/100 m^2"), size = 4)
p <- p + annotate("text", x=7, y=1.15, label=round(herbiv_mean, 0), size = 4)
p <- p + annotate("text", x=7, y=1.2, label=round(herbiv_max, 0), size = 4)
p <- p + annotate("text", x=8, y=1.1, label=paste(round(tempVar_min, 2), "C"), size = 4)
p <- p + annotate("text", x=8, y=1.15, label=paste(round(tempVar_mean, 2), "C"), size = 4)
p <- p + annotate("text", x=8, y=1.2, label=paste(round(tempVar_max, 2), "C"), size = 4)

#Adds annotation saying that summary stats for raw values aren't applicable
for (i in 9:14) {
  
  p <- p + annotate("text", x=i, y=1.15, label="N/A", size=4)

}

p

ggsave("indic_resil_stressor_box_plot.jpg", plot = p, device = "jpeg", width= 15, height=7, dpi=300)

```



Functions for management queries
```{r Functions for management queries}

#Identifies sites that are conducive to coral restoration
#Criteria: 1) hard coral cover is lower than the average across all sites, and 2) site resilience is in the upper two quartiles (excluding the temperature variation indicator)
restorn_axn <- function(indics_stressors, i, cutoff_quartile) {

  #Average hard coral cover across all sites
  coral_cov_avg <- mean(indics_stressors[,c("HardCoralCover_rescaled")], na.rm=TRUE)
    
  #Determines whether each site fits the restoration criteria
  for (j in 1:nrow(indics_stressors)) {
      
    #Applies the two criteria to each site
    if(indics_stressors[j,"HardCoralCover_rescaled"] < coral_cov_avg && indics_stressors[j,"resil_quart_no_temp_var"] <= cutoff_quartile){
        
      #If the site meets both critera, it is labeled "TARGET"
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "restorn_axn"] <- "RESTORE"
        
    } else { 
       
      #If the site does not meet both criteria, it is labeled "NOT TARGET"
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "restorn_axn"] <- "NOT TARGET"
        
    }
  }
    
  return(demog_sites_with_queries)

}


#Identifies sites that are conducive to LBSP reduction
#Criteria: 1) LBSP is higher than the average across all sites, and 2) site resilience is in the upper two quartiles (excluding the temperature variation indicator)
LBSP_axn <- function(indics_stressors, i, cutoff_quartile) {
  
  #Average LBSP values (sediment and nitrogen combined) across all sites that had LBSP values
  LBSP_avg <- mean(indics_stressors[,c("StrdLBSP")], na.rm=TRUE)
  
  #Determines whether each site fits the LBSP management criteria
  for (j in 1:nrow(indics_stressors)) {

    #Marks the site as not having LBSP data
    if(is.na(indics_stressors[j, "StrdLBSP"])) {
      
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "LBSP_axn"] <- "NO LBSP DATA"
      
    } 
    
    #Applies the management criteria to each site
    else if(indics_stressors[j,"StrdLBSP"] > LBSP_avg && indics_stressors[j,"resil_quart_no_temp_var"] <= cutoff_quartile) {

      #If the site meets both critera, it is labeled "TARGET"
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "LBSP_axn"] <- "LBSP"
        
    } else { 
       
      #If the site does not meet both criteria, it is labeled "NOT TARGET"
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "LBSP_axn"] <- "NOT TARGET"
        
    }
  }
    
  return(demog_sites_with_queries)

}


#Identifies sites that are conducive to bleaching management
#Criteria: 1) Bleaching resistance is lower than the average across all sites, and 2) herbivore biomass is lower than average across all sites
bleaching_axn <- function(indics_stressors, i) {

  #Average herbivore biomass across all sites
  HerbivoreBiomass_avg <- mean(indics_stressors[,c("HerbivoreBiomass_rescaled")], na.rm=TRUE)

  #Average coral thermal tolerance across all sites
  ThermalTol_avg <- mean(indics_stressors[,c("ThermalTol_rescaled")], na.rm=TRUE)  
  
  #Determines whether each site fits the bleaching management criteria
  for (j in 1:nrow(indics_stressors)) {

    #Marks the site as not having coral demographic data
    if(is.na(indics_stressors[j, "ThermalTol_rescaled"])) {
      
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "bleaching_axn"] <- "NO DEMOG DATA/NO CORAL COLONIES"

    } 
    
    #Applies the management criteria to each site
    else if(indics_stressors[j,"HerbivoreBiomass_rescaled"] < HerbivoreBiomass_avg && indics_stressors[j,"ThermalTol_rescaled"] < ThermalTol_avg) {

      #If the site meets both critera, it is labeled "TARGET"
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "bleaching_axn"] <- "BLEACHING"
        
    } else { 
       
      #If the site does not meet both criteria, it is labeled "NOT TARGET"
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "bleaching_axn"] <- "NOT TARGET"
        
    }
  }
    
  return(demog_sites_with_queries)

}


#Identifies sites that are conducive to tourism outreach
#Criteria: 1) Coral diversity is higher than average, 2) algae cover is lower than average, and 3) herbivore biomass is higher than average
tourism_axn <- function(indics_stressors, i) {

  #Average coral diversity across all sites
  SimpsonDiv_avg <- mean(indics_stressors[,c("SimpsonDiv_rescaled")], na.rm=TRUE)  
  
  #Average algae cover across all sites
  AlgaeCover_avg <- mean(indics_stressors[,c("AlgaeCover_rescaled")], na.rm=TRUE)  
  
  #Average herbivore biomass across all sites
  HerbivoreBiomass_avg <- mean(indics_stressors[,c("HerbivoreBiomass_rescaled")], na.rm=TRUE)
  
  #Determines whether each site fits the tourism outreach criteria
  for (j in 1:nrow(indics_stressors)) {

    #Marks the site as not having the necessary data
    if(is.na(indics_stressors[j, "SimpsonDiv_rescaled"])) {
      
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "tourism_axn"] <- "NO DEMOG DATA/NO CORAL COLONIES"

    } 
    
    #Applies the management criteria to each site
    else if(indics_stressors[j,"SimpsonDiv_rescaled"] > SimpsonDiv_avg && indics_stressors[j,"AlgaeCover_rescaled"] < AlgaeCover_avg & indics_stressors[j,"HerbivoreBiomass_rescaled"] >  HerbivoreBiomass_avg) {

      #If the site meets all critera, it is labeled "TARGET"
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "tourism_axn"] <- "TOURISM"
        
    } else { 
       
      #If the site does not meet both criteria, it is labeled "NOT TARGET"
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "tourism_axn"] <- "NOT TARGET"
        
    }
  }
    
  return(demog_sites_with_queries)

}


#Identifies sites that are conducive to fishery management and enforcement
#Criteria: 1) Herbivore biomass is lower than the average across all sites, and 2) fishing pressure is higher than average across all sites
fishing_axn <- function(indics_stressors, i) {

  #Average herbivore biomass across all sites
  HerbivoreBiomass_avg <- mean(indics_stressors[,c("HerbivoreBiomass_rescaled")], na.rm=TRUE)

  #Average fishing pressure across all sites
  fishing_avg <- mean(indics_stressors[,c("all_total_fishing_rescaled")], na.rm=TRUE)  
  
  #Determines whether each site fits the fishing management criteria
  for (j in 1:nrow(indics_stressors)) {
      
    #Applies the two criteria to each site
    if(indics_stressors[j,"HerbivoreBiomass_rescaled"] < HerbivoreBiomass_avg && indics_stressors[j,"all_total_fishing_rescaled"] > fishing_avg){
        
      #If the site meets both critera, it is labeled "TARGET"
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "fishing_axn"] <- "FISHING"
        
    } else { 
       
      #If the site does not meet both criteria, it is labeled "NOT TARGET"
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "fishing_axn"] <- "NOT TARGET"
        
    }
  }
    
  return(demog_sites_with_queries)

}


#Identifies sites that are conducive to disease monitoring
#Criteria: 1) Coral cover is higher than average across all sites, and 2) disease prevalence is higher than the average across all sites
disease_axn <- function(indics_stressors, i) {

  #Average coral cover across all sites
  CoralCover_avg <- mean(indics_stressors[,c("HardCoralCover_rescaled")], na.rm=TRUE)

  #Average fraction of colonies without disease across all sites
  NotDiseased_avg <- mean(indics_stressors[,c("FractNotDiseased_rescaled")], na.rm=TRUE)  
  
  #Determines whether each site fits the disease management criteria
  for (j in 1:nrow(indics_stressors)) {

        #Marks the site as not having the necessary data
    if(is.na(indics_stressors[j, "FractNotDiseased_rescaled"])) {
      
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "disease_axn"] <- "NO DEMOG DATA/NO CORAL COLONIES"

    } 
     
    #Applies the two criteria to each site
    else if(indics_stressors[j,"HardCoralCover_rescaled"] > CoralCover_avg && indics_stressors[j,"FractNotDiseased_rescaled"] < NotDiseased_avg){

      #If the site meets both critera, it is labeled "TARGET"
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "disease_axn"] <- "DISEASE"
        
    } else { 
       
      #If the site does not meet both criteria, it is labeled "NOT TARGET"
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "disease_axn"] <- "NOT TARGET"
        
    }
  }
    
  return(demog_sites_with_queries)

}
 

#Identifies sites that are conducive to coral protection
#Criteria: 1) LBSP is moderate, 2) fishing is moderate, and 3) resilience is in the upper two quartiles (excluding the temperature variation indicator)
protexn_axn <- function(indics_stressors, i, cutoff_quartile) {

  #Average and standard deviation of LBSP values (sediment and nitrogen combined) across all sites that had LBSP values
  LBSP_avg <- mean(indics_stressors[,c("StrdLBSP")], na.rm=TRUE)
  LBSP_stdev <- sd(indics_stressors[,c("StrdLBSP")], na.rm=TRUE)

  #Average and standard deviation of fishing pressure across all sites
  fishing_avg <- mean(indics_stressors[,c("all_total_fishing_rescaled")], na.rm=TRUE) 
  fishing_stdev <- sd(indics_stressors[,c("all_total_fishing_rescaled")], na.rm=TRUE)

  #Determines whether each site fits the protection criteria
  for (j in 1:nrow(indics_stressors)) {

    #Marks the site as not having LBSP data
    if(is.na(indics_stressors[j, "StrdLBSP"])) {
      
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "protexn_axn"] <- "NO LBSP DATA"
      
    } 
    
    #Applies the management criteria to each site
    else if(indics_stressors[j,"StrdLBSP"] <= LBSP_avg + LBSP_stdev && indics_stressors[j,"StrdLBSP"] >= LBSP_avg - LBSP_stdev && indics_stressors[j,"all_total_fishing_rescaled"] <= fishing_avg + fishing_stdev && indics_stressors[j,"all_total_fishing_rescaled"] >= fishing_avg - fishing_stdev && indics_stressors[j,"resil_quart_no_temp_var"] <= cutoff_quartile) {

      #If the site meets both critera, it is labeled "TARGET"
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "protexn_axn"] <- "PROTECT"
        
    } else { 
       
      #If the site does not meet both criteria, it is labeled "NOT TARGET"
      demog_sites_with_queries[i*nrow(indics_stressors) + j, "protexn_axn"] <- "NOT TARGET"
        
    }
  }
    
  return(demog_sites_with_queries)

}


```



Runs management queries using indicators, resilience scores, and fishing/LBSP stressor values
```{r Runs management queries using indicators, resilience scores, and fish/LBSP stressor values}

#Dataframe that stores the management query results
demog_sites_with_queries <- demog_sites_with_stressors

#The number of higher resilience quartiles which are being used for querying management
#e.g., "2" means that sites in the two highest quartiles will be considered for the management actions
cutoff_quartile <- 2

#Column that stores whether each site is a restoration target
demog_sites_with_queries$restorn_axn <- NA

#Column that stores whether each site is a LBSP reduction target
demog_sites_with_queries$LBSP_axn <- NA

#Column that stores whether each site is a bleaching management target
demog_sites_with_queries$bleaching_axn <- NA

#Column that stores whether each site is a tourism outreach target
demog_sites_with_queries$tourism_axn <- NA

#Column that stores whether each site is a fishing management target
demog_sites_with_queries$fishing_axn <- NA

#Column that stores whether each site is a coral disease monitoring and management target
demog_sites_with_queries$disease_axn <- NA

#Column that stores whether each site is a protection target
demog_sites_with_queries$protexn_axn <- NA

#Identifies sites that meet the management action query criteria. Uses resilience scores without temperature variation indicator, wherever applicable.
#Iterates through each weighting system.
for (i in 0:max(demog_sites_with_queries$weighting_sys)) {
   
  #Determines whether each site is a restoration target under each weighting system 
  demog_sites_with_queries <- restorn_axn(demog_sites_with_queries[which(demog_sites_with_queries$weighting_sys == i),], i, cutoff_quartile)
  
  #Determines whether each site is an LBSP management target under each weighting system
  demog_sites_with_queries <- LBSP_axn(demog_sites_with_queries[which(demog_sites_with_queries$weighting_sys == i),], i, cutoff_quartile)
  
  #Determines whether each site is a bleaching monitoring and management target under each weighting system
  demog_sites_with_queries <- bleaching_axn(demog_sites_with_queries[which(demog_sites_with_queries$weighting_sys == i),], i)
  
  #Determines whether each site would be a good target for tourism outreach
  demog_sites_with_queries <- tourism_axn(demog_sites_with_queries[which(demog_sites_with_queries$weighting_sys == i),], i)
  
  #Determines whether each site would be a good target for fishing management
  demog_sites_with_queries <- fishing_axn(demog_sites_with_queries[which(demog_sites_with_queries$weighting_sys == i),], i)
  
  #Determines whether each site would be a good target for disease management
  demog_sites_with_queries <- disease_axn(demog_sites_with_queries[which(demog_sites_with_queries$weighting_sys == i),], i)
  
  #Determines whether each site is a protection target under each weighting system
  demog_sites_with_queries <- protexn_axn(demog_sites_with_queries[which(demog_sites_with_queries$weighting_sys == i),], i, cutoff_quartile)
  
}

#Column that stores all of the actions that can be used at that site
demog_sites_with_queries$all_axns <- apply(demog_sites_with_queries[,c(68:74)], 1, paste, collapse= " ")

#Removes the text saying which actions don't apply; leaves only which actions do apply at that site
demog_sites_with_queries$all_axns <- gsub("NOT TARGET ", "", demog_sites_with_queries$all_axns)
demog_sites_with_queries$all_axns <- gsub("NO DEMOG DATA/NO CORAL COLONIES ", "", demog_sites_with_queries$all_axns)
demog_sites_with_queries$all_axns <- gsub("NOT TARGET", "", demog_sites_with_queries$all_axns)
demog_sites_with_queries$all_axns <- gsub("NO LBSP DATA ", "", demog_sites_with_queries$all_axns)
demog_sites_with_queries$all_axns <- gsub("NO LBSP DATA", "", demog_sites_with_queries$all_axns)
demog_sites_with_queries$all_axns <- gsub(" $", "", demog_sites_with_queries$all_axns)
demog_sites_with_queries$all_axns <- gsub("^$", "NO_ACTION", demog_sites_with_queries$all_axns)

#Column that stores just the first letter or each action
demog_sites_with_queries$all_axns_brief <- demog_sites_with_queries$all_axns

#Supplies the first letter of each action
demog_sites_with_queries$all_axns_brief <- gsub("RESTORE", "R", demog_sites_with_queries$all_axns_brief)
demog_sites_with_queries$all_axns_brief <- gsub("LBSP", "L", demog_sites_with_queries$all_axns_brief)
demog_sites_with_queries$all_axns_brief <- gsub("BLEACHING", "B", demog_sites_with_queries$all_axns_brief)
demog_sites_with_queries$all_axns_brief <- gsub("TOURISM", "T", demog_sites_with_queries$all_axns_brief)
demog_sites_with_queries$all_axns_brief <- gsub("FISHING", "F", demog_sites_with_queries$all_axns_brief)
demog_sites_with_queries$all_axns_brief <- gsub("DISEASE", "D", demog_sites_with_queries$all_axns_brief)
demog_sites_with_queries$all_axns_brief <- gsub("PROTECT", "P", demog_sites_with_queries$all_axns_brief)
demog_sites_with_queries$all_axns_brief <- gsub(" ", "", demog_sites_with_queries$all_axns_brief)


write.table(demog_sites_with_queries, file = paste("resil_with_queries_", Sys.Date(), ".csv", sep=""), row.names = FALSE)

```



```{r Adds the year of onset of annual bleaching under RCP8.5}

demog_sites_with_bleaching <- merge(demog_sites_with_queries, annual_bleaching_rcp85[,c("Station", "onset_yr_annual_bleaching_rcp85", "outside_annual_bleaching_proj_raster")], by.x = "Station", by.y = "Station")

#Since the above merges changed the order of the rows, reorders the indicator/stressor data by the weighting system, then by the station ID
demog_sites_with_bleaching <- demog_sites_with_bleaching[order(demog_sites_with_bleaching$weighting_sys, demog_sites_with_bleaching$Station),]

#Field to categorize sites by their temperature exposure (annual significant bleaching onset year)
demog_sites_with_bleaching$exposure <- NA

#Field to store vulnerablity of sites (combination of exposure and resilience)
demog_sites_with_bleaching$vulnerability <- NA

#Iterates through every row
for(i in 1:nrow(demog_sites_with_bleaching)) {
  
  #Categorizes sites by their temperature exposure
  if(demog_sites_with_bleaching[i,"onset_yr_annual_bleaching_rcp85"] < 2042) {
    demog_sites_with_bleaching[i,"exposure"] <- "HIGH"
  }
  else {
    demog_sites_with_bleaching[i,"exposure"] <- "LOW"
  }
  
  #Categorizes each row into vulnerability levels (low, moderate, high) based on temperature projection exposure and resilience quartile (without temperature variation indicator)
  if(demog_sites_with_bleaching[i,"exposure"] == "LOW" & demog_sites_with_bleaching[i,"resil_quart_no_temp_var"] < 3) {
    demog_sites_with_bleaching[i,"vulnerability"] <- "LOW"
  }
  else if(demog_sites_with_bleaching[i,"exposure"] == "LOW" && demog_sites_with_bleaching[i,"resil_quart_no_temp_var"] > 2) {
    demog_sites_with_bleaching[i,"vulnerability"] <- "MODERATE"
  }
  else if(demog_sites_with_bleaching[i,"exposure"] == "HIGH" && demog_sites_with_bleaching[i,"resil_quart_no_temp_var"] < 3) {
    demog_sites_with_bleaching[i,"vulnerability"] <- "MODERATE"
  }
  else {
    demog_sites_with_bleaching[i,"vulnerability"] <- "HIGH"
  }
  
}


write.table(demog_sites_with_bleaching, file = paste("resil_with_bleaching_onset_", Sys.Date(), ".csv", sep=""), row.names = FALSE)

```




Miscellaneous calculations
```{r Miscellaneous calculations}

#Standard deviations of resilience scores with and without temperature variation indicator
sd(data_demog_sites_unweighted$resil_all_indic)
sd(data_demog_sites_unweighted$resil_no_temp_var)

#Correlation between resilience scores with and without temperature variation indicator
cor(data_demog_sites_unweighted$resil_all_indic, data_demog_sites_unweighted$resil_no_temp_var, method="pearson")

```



Canonical analysis of principal coordinates
From BiodiverstiyR package. Used by Maynard et al. 2015 for CNMI report and described in Maynard et al. 2017 UN report
```{r Canonical analysis of principal coordinates (CAP)}

#Data.frame of rescaled indicators
indics <- data_demog_sites_unweighted[,c(10:17)]
indics_df <- as.data.frame(indics)

#the variable being classified by is resilience quartile without temperature variation
indics.env <- data_demog_sites_unweighted[,c(43:46)]
indics.env[,4] <- as.factor(indics.env[,4])

#Runs the model
indic_model <- CAPdiscrim(indics_df~resil_quart_no_temp_var, data=indics.env, dist="bray", axes=2, m=0, add=FALSE)
indic_model

colvec <- c("green4", "orange", "yellow", "red")

#Plots the CAP groupings
indic_plot <- ordiplot(indic_model, type="none", xlab="CAP1", ylab="CAP2", col=colvec)
ordisymbol(indic_plot, indics.env, "resil_quart_no_temp_var", legend=TRUE)

ordiellipse(indic_model, indics.env$resil_quart_no_temp_var, scaling="symmetric", col=1:4, draw="polygon", label=TRUE)



```